陈亨,邓非

• 1:自然资源部城市国土资源监测与仿真重点实验室
• 2:武汉大学

摘要(Abstract)：

针对三维卷积神经网络(3D-CNN)计算成本过大,训练、测试时间较长的问题,该文提出了一种分解式三维卷积神经网络(Dec-3D-CNN)。通过将一步三维卷积运算拆分成三步更简单的卷积运算来降低计算成本,并且结合批量标准化(BN)的方法共同设计神经网络结构。在加速网络训练的同时减少梯度弥散的情况。通过Pavia University数据集进行分类实验,Dec-3D-CNN在总体分类精度达到95.93%的情况下,训练时间仅为3D-CNN的16%,测试时间仅为3D-CNN的46%。实验结果表明,Dec-3D-CNN在保持高精度的情况下,能够大幅度的节省训练时间,降低计算成本。

关键词(KeyWords)： 高光谱遥感影像分类;三维卷积;批量标准化;支持向量机;K-最近邻分类

Abstract:

Keywords:

基金项目(Foundation): 自然资源部城市国土资源监测与仿真重点实验室开放基金资助课题项目(KF-2018-03-025);; 战略火箭创新基金项目

作者(Author): 陈亨,邓非

DOI: 10.16251/j.cnki.1009-2307.2020.08.015

参考文献(References)：

• [1] GUO Yiqiang,WU Yanbin,JU Zhengshan,et al.Remote sensing image classification by the Chaos genetic algorithm in monitoring land use changes[J].Mathematical and Computer Modelling,2010,51(11/12):1408-1416.
• [2] DUVEILLER G,DEFOURNY P.A conceptual framework to define the spatial resolution requirements for agricultural monitoring using remote sensing[J].Remote Sensing of Environment,2010,114(11):2637-2650.
• [3] LI X,DAMEN M C J.Coastline change detection with satellite remote sensing for environmental management of the Pearl River Estuary,China[J].Journal of Marine Systems,2010,82(S):54-61.
• [4] ZHAO C,LIU W,XU Y,et al.A spectral-spatial SVM-based multi-layer learning algorithm for hyperspectral image classification[J].Remote Sensing Letters,2018,9(3):218-227.
• [5] XIA J,FALCO N,BENEDIKTSSON J A,et al.Hyperspectral image classification with rotation random forest via KPCA[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2017,10(4):1601-1609.
• [6] WANG L Z,HUANG H,FENG H L.Hyperspectral remote sensing image classification based on MFA and KNNS[J].Computer Science,2012,40(4):780-787.
• [7] 王增茂,杜博,张良培,等.基于纹理特征和形态学特征融合的高光谱影像分类法[J].光子学报,2014,43(8):122-129.(WANG Zengmao,DU Bo,ZHANG Liangpei,et al.Based on texture feature and extend morphological profile fusion for hyperspectral image classification[J].Acta Photonica Sinica,2014,43(8):122-129.)
• [8] KUMAR B,DIKSHIT O.Hyperspectral image classification based on morphological profiles and decision fusion[M].Abingdon:Taylor & Francis,Inc,2017.
• [9] 曹林林,李海涛,韩颜顺,等.卷积神经网络在高分遥感影像分类中的应用[J].测绘科学,2016,41(9):170-175.(CAO Linlin,LI Haitao,HAN Yanshun,et al.Application of convolutional neural networks in classification of high resolution remote sensing imagery[J].Science of Surveying and Mapping,2016,41(9):170-175.)
• [10] SZEGEDY C,IOFFE S,VANHOUCKE V,et al.Inception-v4,Inception-ResNet and the impact of residual connections on learning[J/OL].[2019-04-03].https://arxiv.org/abs/1602.07261.
• [11] LONG J,SHELHAMER E,DARRELL T.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2014,39(4):640-651.
• [12] MOLCHANOV P,YANG X,GUPTA S,et al.Online detection and classification of dynamic hand gestures with recurrent 3D convolutional neural networks[C]//Computer Vision & Pattern Recognition.[S.l.]:[s.n.],2016.
• [13] LI Y,ZHANG H K,SHEN Q.Spectral-spatial classification of hyperspectral imagery with 3D convolutional neural network[J].Remote Sensing,2017,9(1):67.
• [14]LENG J B,LI T,BAI G,et al.Cube-CNN-SVM:a novel hyperspectral image classification method[C]//IEEE International Conference on Tools with Artificial Intelligence.[S.l.]:IEEE,2017.
• [15]HOWARD A G,ZHU M L, CHEN B,et al.MobileNets:efficient convolutional neural networks for mobile vision applications[J/OL].[2019-04-03].https://arxiv.org/pdf/1704.04861.pdf.
• [16]IOFFE S,SZEGEDY C.Batch normalization:accelerating deep network training by reducing internal covariate shift[J/OL].[2019-04-03].https://arxiv.org/abs/1502.03167.
• [17]SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al. Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
• [18]HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition.[S.l.]:IEEE,2015.
• [19]ABADI M,BARHAM P,CHEN J M,et al.TensorFlow:a system for large-scale machine learning[J/OL].[2019-04-03].https://arxiv.org/abs/1605.08695.